A Non-Linear Model for Sensing Moisture Content in Transformers at Reduced Time

Abstract

Moisture content in transformer is a prime indicator of its insulation condition. The presence of moisture accelerates the aging of cellulosic paper insulation thereby reducing the dielectric as well as the mechanical strength of the insulation. In the conventional method for sensing moisture content in transformers, sinusoidal excitation is generally applied on the oil-paper insulation system for two cycles from 0.1 mHz to 1 kHz. However, the application of sinusoidal excitations at discrete measurement frequencies leads to long measurement time during low frequency measurement (from 0.1 mHz to 1 Hz), which is a practical problem. Besides, as the moisture content increases, the transformer insulation no longer behaves as a linear time-invariant system. Therefore, non-linear modeling of insulation is necessary for accurate moisture sensing in transformers. Considering these two facts, in this study a non-linear Wiener model of transformer insulation using sine sweep excitation is proposed for reliable sensing of moisture content at reduced measurement time. It was observed that the proposed non-linear Wiener model has better moisture-sensing capability, thereby providing more accurate information regarding the insulation condition in comparison with the linear model. Moreover, an appreciable reduction in the total measurement duration was observed, enabling transformer diagnosis to be conducted quickly compared to the conventional method.